CLM and ILM

[rooble]

Dear experts,


Please enlight me about CLM & ILM.

1. Wht is ILM and why it is used?
2. Wht is CLM and why it is used?

Thanks and regards,

Rahul

[g3jza · Accepted solution]

Hi,
On cell-based servers that are based on the HP sx1000 or sx2000 chipset, a portion of the memory in each cell can be configured as cell local memory (CLM), which is non-interleaved memory that can be quickly accessed by processors residing on the same cell as the memory.

Opposite to that is ILM (interleaved memory) , which is type of memory that can be accessed by processors residing on different cells in nPar as the memory.

There are certain recommendations for vPars for example, which define (best practices) what should be the ratio in the cell for CLM:ILM .

[rooble]

Hi experts,

If ILM becomes faulty than more than one cll will affact. and If CLM becomes faulty than only local cell will affect.

Minium memory must be there to bring a cell up doesn't matter its CLM or ILM is it?

Please correct me if I am wrong.

Regards,

Rahul

[g3jza]

In HP-UX 11.11 CLM is not supported, so the cells belonging to the nPar running 11.11 need to have 100% ILM memory.

CLM is being used in 11.23 + 11.31.

The optimal CLM settings depend on the applications and workload the OS is running.

[g3jza]

One more thing, some other Operating systems don't support CLM at all, so you need to check this first, if you plan on running non-HPUX OS on those cells.

And answering your questions regarding which cells will be affected during failure when using ILM / CLM , I'm not sure that it is exactly like this. Maybe someone here confirms that.

[Emil Velez]

The technical paper enclosed discusses it was well as process memory enjoy

[Ismail Azad]

Hi robbie,

When dealing with concepts of CLM and ILM pops up the *awe inspiring* term ccNUMA. A processor normally can access memory anywhere on the server and in cell based systems shows up the inevitable effect of non uniformity. A processor accessing memory on the same cell takes way less time. A processor accessing memory on another cell on the same cross bar would take a little more time. A processor accessing memory on the same cell on a different crossbar takes way more time.

And hitting with a result of non uniformity in the sense of latency and bandwidth.... *ccNUMA* which has now become a significant feature... Don't forget to assign points.

Regards
Ismail Azad